The present invention relates to the field of color reproduction and, more particularly, to determining true surface color of a surface from multiple camera images providing conflicting color information.
Cameras with multiple fixed lenses are increasingly being utilized to capture imagery in an arc which is greater than what a single lens is able to provide. For any given point, however, the different camera lenses record different color values. These differences are primarily a result of specular reflection. Semi-reflective surfaces are of particular concern for color accuracy. For example, if a red semi-reflective surface is fifty percent diffuse and fifty percent reflective and a light source is white, then color of the surface as captured by a camera lens will appear from red to white color depending on viewing angle and the intensity of the white light source.
Differences in color from different images must be resolved if the images are to be stitched together to form a cohesive scheme. When combining content from different images into a single one, conventional techniques generally try to smooth the color differences, which results in a cohesive combined image. The resulting image, however, does not accurately represent the true color of the object being captured. For semi-reflective surfaces, color accuracy resulting from conventional techniques can be particularly poor. Stated differently, conventional techniques for combining points captured by multiple lenses fail to produce photorealistic images. Which is a problem heretofore unresolved.
Additionally, many image analysis techniques assume an object's true color is represented. These techniques yield inaccurate results when this assumption is incorrect. For instance, color of an image is often used to determine a shape of a reconstructed surface. The more accurate the color; the more accurate the determined shape. Embodiments of the disclosure ensure true color of surfaces is maintained when new images are being generated by stitching together a set of images captured using different lenses positioned at different angles relative to a surface.